Scour around bridge piers poses significant risks in hydraulic engineering, with traditional empirical models often lacking the precision to capture complex interactions. This study employs a machine learning approach, combining Random Forest (RF) and Artificial Neural Networks (ANN), to enhance scour prediction accuracy. Using a dataset of 208 observations, RF identified key factors such as flow intensity (V/Vc), pier form (Dp/Y), and sediment gradation (σ). ANN models, particularly with the ReLU activation function, demonstrated superior performance, achieving an R2 of 0.92 for training and 0.76 for testing. These findings recommend machine learning techniques for improving scour risk assessments and mitigation strategies.